Attribution de l'incertitude structurelle du modèle hydrologique dans les études d'impact du changement climatique

Abstract

This research investigated the importance of the structure of hydrological models as a source of uncertainty in studies of the impacts of climate change on water resources. Several recent studies have shown that structural uncertainty can play a significant and sometimes even dominant role on future uncertainty, but the origin of this uncertainty has been little explored. In order to explore the uncertainty associated with the structure of the hydrological model and to understand the uncertainty associated with each of the structural components, a multi-model analysis was performed using eleven (11) representations of evapotranspiration, six (6) models of snow accumulation and melt and sixteen (16) formulations of water flow forming 1056 combinations of hydrological structures. The study covers 21 hydrological metrics and a set of 100 watersheds located in North America. The importance of these three components in generating uncertainty in present and future climates was investigated and compared with that related to general circulation models. The results showed that for the present climate, the structural uncertainty resides mostly in the water flow formulations for all metrics considered. In a warmer future climate, the structural uncertainty in the hydrological models continues to dominate the low flow and large extreme flow metrics and is otherwise comparable to that resulting from the climate models. Potential evapotranspiration formulations in the future climate contribute more strongly to the overall uncertainty and is comparable to that from runoff formulations for metrics describing medium to high flows but becomes minimal for metrics representing low and very high flows. The contribution to variance from snow accumulation and melt formulations remains small for all cases and metrics considered. A selection of the best hydrological model structures has relatively little impact on the overall and relative uncertainties. It reduces the contribution of high order interaction terms related to extreme flows (return period of 2 to 20 years), and enhances the relative contribution of GCMs, but does not significantly reduce the total variance, or drastically reduce the contribution of hydrological structures to the total variance.

Date	1 Nov 2021
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	François Brissette (Supervisor)